Abstract: BackgroundThe timing of renal-replacement therapy in critically ill patients who have acute kidney injury but no potentially life-threatening complication directly related to renal failure is a subject of debate. MethodsIn this multicenter randomized trial, we assigned patients with severe acute kidney injury (Kidney Disease: Improving Global Outcomes [KDIGO] classification, stage 3 [stages range from 1 to 3, with higher stages indicating more severe kidney injury]) who required mechanical ventilation, catecholamine infusion, or both and did not have a potentially life-threatening complication directly related to renal failure to either an early or a delayed strategy of renal-replacement therapy. With the early strategy, renal-replacement therapy was started immediately after randomization. With the delayed strategy, renal-replacement therapy was initiated if at least one of the following criteria was met: severe hyperkalemia, metabolic acidosis, pulmonary edema, blood urea nitrogen level higher than 112 ...

Abstract: The security of computer systems fundamentally relies on memory isolation, e.g., kernel address ranges are marked as non-accessible and are protected from user access. In this paper, we present Meltdown. Meltdown exploits side effects of out-of-order execution on modern processors to read arbitrary kernel-memory locations including personal data and passwords. Out-of-order execution is an indispensable performance feature and present in a wide range of modern processors. The attack is independent of the operating system, and it does not rely on any software vulnerabilities. Meltdown breaks all security guarantees provided by address space isolation as well as paravirtualized environments and, thus, every security mechanism building upon this foundation. On affected systems, Meltdown enables an adversary to read memory of other processes or virtual machines in the cloud without any permissions or privileges, affecting millions of customers and virtually every user of a personal computer. We show that the KAISER defense mechanism for KASLR has the important (but inadvertent) side effect of impeding Meltdown. We stress that KAISER must be deployed immediately to prevent large-scale exploitation of this severe information leakage.

Abstract: Immune escape is a prerequisite for tumor development. To avoid the immune system, tumors develop different mechanisms, including T cell exhaustion, which is characterized by expression of immune inhibitory receptors, such as PD-1, CTLA-4, Tim-3, and a progressive loss of function. The recent development of therapies targeting PD-1 and CTLA-4 have raised great interest since they induced long-lasting objective responses in patients suffering from advanced metastatic tumors. However, the regulation of PD-1 expression, and thereby of exhaustion, is unclear. VEGF-A, a proangiogenic molecule produced by the tumors, plays a key role in the development of an immunosuppressive microenvironment. We report in the present work that VEGF-A produced in the tumor microenvironment enhances expression of PD-1 and other inhibitory checkpoints involved in CD8+ T cell exhaustion, which could be reverted by anti-angiogenic agents targeting VEGF-A–VEGFR. In view of these results, association of anti-angiogenic molecules with immunomodulators of inhibitory checkpoints may be of particular interest in VEGF-A-producing tumors.

Abstract: Purpose:To describe image artifacts of optical coherence tomography (OCT) angiography and their underlying causative mechanisms. To establish a common vocabulary for the artifacts observed.Methods:The methods by which OCT angiography images are acquired, generated, and displayed are reviewed as are

Abstract: Objective Electroencephalography (EEG) analysis has been an important tool in neuroscience with applications in neuroscience, neural engineering (e.g. Brain-computer interfaces, BCI's), and even commercial applications. Many of the analytical tools used in EEG studies have used machine learning to uncover relevant information for neural classification and neuroimaging. Recently, the availability of large EEG data sets and advances in machine learning have both led to the deployment of deep learning architectures, especially in the analysis of EEG signals and in understanding the information it may contain for brain functionality. The robust automatic classification of these signals is an important step towards making the use of EEG more practical in many applications and less reliant on trained professionals. Towards this goal, a systematic review of the literature on deep learning applications to EEG classification was performed to address the following critical questions: (1) Which EEG classification tasks have been explored with deep learning? (2) What input formulations have been used for training the deep networks? (3) Are there specific deep learning network structures suitable for specific types of tasks? Approach A systematic literature review of EEG classification using deep learning was performed on Web of Science and PubMed databases, resulting in 90 identified studies. Those studies were analyzed based on type of task, EEG preprocessing methods, input type, and deep learning architecture. Main results For EEG classification tasks, convolutional neural networks, recurrent neural networks, deep belief networks outperform stacked auto-encoders and multi-layer perceptron neural networks in classification accuracy. The tasks that used deep learning fell into five general groups: emotion recognition, motor imagery, mental workload, seizure detection, event related potential detection, and sleep scoring. For each type of task, we describe the specific input formulation, major characteristics, and end classifier recommendations found through this review. Significance This review summarizes the current practices and performance outcomes in the use of deep learning for EEG classification. Practical suggestions on the selection of many hyperparameters are provided in the hope that they will promote or guide the deployment of deep learning to EEG datasets in future research.

Abstract: The goal of this paper is to generate high-quality 3D object proposals in the context of autonomous driving. Our method exploits stereo imagery to place proposals in the form of 3D bounding boxes. We formulate the problem as minimizing an energy function encoding object size priors, ground plane as well as several depth informed features that reason about free space, point cloud densities and distance to the ground. Our experiments show significant performance gains over existing RGB and RGB-D object proposal methods on the challenging KITTI benchmark. Combined with convolutional neural net (CNN) scoring, our approach outperforms all existing results on all three KITTI object classes.

Abstract: We train a generative convolutional neural network which is able to generate images of objects given object type, viewpoint, and color. We train the network in a supervised manner on a dataset of rendered 3D chair models. Our experiments show that the network does not merely learn all images by heart, but rather finds a meaningful representation of a 3D chair model allowing it to assess the similarity of different chairs, interpolate between given viewpoints to generate the missing ones, or invent new chair styles by interpolating between chairs from the training set. We show that the network can be used to find correspondences between different chairs from the dataset, outperforming existing approaches on this task.

Abstract: Electrochemical analysis of sweat using soft bioelectronics on human skin provides a new route for noninvasive glucose monitoring without painful blood collection. However, sweat-based glucose sensing still faces many challenges, such as difficulty in sweat collection, activity variation of glucose oxidase due to lactic acid secretion and ambient temperature changes, and delamination of the enzyme when exposed to mechanical friction and skin deformation. Precise point-of-care therapy in response to the measured glucose levels is still very challenging. We present a wearable/disposable sweat-based glucose monitoring device integrated with a feedback transdermal drug delivery module. Careful multilayer patch design and miniaturization of sensors increase the efficiency of the sweat collection and sensing process. Multimodal glucose sensing, as well as its real-time correction based on pH, temperature, and humidity measurements, maximizes the accuracy of the sensing. The minimal layout design of the same sensors also enables a strip-type disposable device. Drugs for the feedback transdermal therapy are loaded on two different temperature-responsive phase change nanoparticles. These nanoparticles are embedded in hyaluronic acid hydrogel microneedles, which are additionally coated with phase change materials. This enables multistage, spatially patterned, and precisely controlled drug release in response to the patient’s glucose level. The system provides a novel closed-loop solution for the noninvasive sweat-based management of diabetes mellitus.

Abstract: While the patient and public involvement (PPI) evidence base has expanded over the past decade, the quality of reporting within papers is often inconsistent, limiting our understanding of how it works, in what context, for whom, and why. To develop international consensus on the key items to report to enhance the quality, transparency, and consistency of the PPI evidence base. To collaboratively involve patients as research partners at all stages in the development of GRIPP2. The EQUATOR method for developing reporting guidelines was used. The original GRIPP (Guidance for Reporting Involvement of Patients and the Public) checklist was revised, based on updated systematic review evidence. A three round Delphi survey was used to develop consensus on items to be included in the guideline. A subsequent face-to-face meeting produced agreement on items not reaching consensus during the Delphi process. One hundred forty-three participants agreed to participate in round one, with an 86% (123/143) response for round two and a 78% (112/143) response for round three. The Delphi survey identified the need for long form (LF) and short form (SF) versions. GRIPP2-LF includes 34 items on aims, definitions, concepts and theory, methods, stages and nature of involvement, context, capture or measurement of impact, outcomes, economic assessment, and reflections and is suitable for studies where the main focus is PPI. GRIPP2-SF includes five items on aims, methods, results, outcomes, and critical perspective and is suitable for studies where PPI is a secondary focus. GRIPP2-LF and GRIPP2-SF represent the first international evidence based, consensus informed guidance for reporting patient and public involvement in research. Both versions of GRIPP2 aim to improve the quality, transparency, and consistency of the international PPI evidence base, to ensure PPI practice is based on the best evidence. In order to encourage its wide dissemination this article is freely accessible on The BMJ and Research Involvement and Engagement journal websites.

Abstract: Modern computer vision algorithms typically require expensive data acquisition and accurate manual labeling. In this work, we instead leverage the recent progress in computer graphics to generate fully labeled, dynamic, and photo-realistic proxy virtual worlds. We propose an efficient real-to-virtual world cloning method, and validate our approach by building and publicly releasing a new video dataset, called Virtual KITTI (see this http URL), automatically labeled with accurate ground truth for object detection, tracking, scene and instance segmentation, depth, and optical flow. We provide quantitative experimental evidence suggesting that (i) modern deep learning algorithms pre-trained on real data behave similarly in real and virtual worlds, and (ii) pre-training on virtual data improves performance. As the gap between real and virtual worlds is small, virtual worlds enable measuring the impact of various weather and imaging conditions on recognition performance, all other things being equal. We show these factors may affect drastically otherwise high-performing deep models for tracking.

Abstract: We present a new approach for the approximate K-nearest neighbor search based on navigable small world graphs with controllable hierarchy (Hierarchical NSW, HNSW). The proposed solution is fully graph-based, without any need for additional search structures (typically used at the coarse search stage of the most proximity graph techniques). Hierarchical NSW incrementally builds a multi-layer structure consisting of a hierarchical set of proximity graphs (layers) for nested subsets of the stored elements. The maximum layer in which an element is present is selected randomly with an exponentially decaying probability distribution. This allows producing graphs similar to the previously studied Navigable Small World (NSW) structures while additionally having the links separated by their characteristic distance scales. Starting the search from the upper layer together with utilizing the scale separation boosts the performance compared to NSW and allows a logarithmic complexity scaling. Additional employment of a heuristic for selecting proximity graph neighbors significantly increases performance at high recall and in case of highly clustered data. Performance evaluation has demonstrated that the proposed general metric space search index is able to strongly outperform previous opensource state-of-the-art vector-only approaches. Similarity of the algorithm to the skip list structure allows straightforward balanced distributed implementation.

Abstract: Background Idecabtagene vicleucel (ide-cel, also called bb2121), a B-cell maturation antigen–directed chimeric antigen receptor (CAR) T-cell therapy, has shown clinical activity with expec...

Abstract: A newly identified coronavirus, 2019-nCoV, has been posing significant threats to public health since December 2019. ACE2, the host cell receptor for severe acute respiratory syndrome coronavirus (SARS), has recently been demonstrated in mediating 2019-nCoV infection. Interestingly, besides the respiratory system, substantial proportion of SARS and 2019-nCoV patients showed signs of various degrees of liver damage, the mechanism and implication of which have not yet been determined. Here, we performed an unbiased evaluation of cell type specific expression of ACE2 in healthy liver tissues using single cell RNA-seq data of two independent cohorts, and identified specific expression in cholangiocytes. The results indicated that virus might directly bind to ACE2 positive cholangiocytes but not necessarily hepatocytes. This finding suggested the liver abnormalities of SARS and 2019-nCoV patients may not be due to hepatocyte damage, but cholangiocyte dysfunction and other causes such as drug induced and systemic inflammatory response induced liver injury. Our findings indicate that special care of liver dysfunction should be installed in treating 2019-nCoV patients during the hospitalization and shortly after cure.

Abstract: Trusted execution environments, and particularly the Software Guard eXtensions (SGX) included in recent Intel x86 processors, gained significant traction in recent years. A long track of research papers, and increasingly also real-world industry applications, take advantage of the strong hardware-enforced confidentiality and integrity guarantees provided by Intel SGX. Ultimately, enclaved execution holds the compelling potential of securely offloading sensitive computations to untrusted remote platforms. We present Foreshadow, a practical software-only microarchitectural attack that decisively dismantles the security objectives of current SGX implementations. Crucially, unlike previous SGX attacks, we do not make any assumptions on the victim enclave's code and do not necessarily require kernel-level access. At its core, Foreshadow abuses a speculative execution bug in modern Intel processors, on top of which we develop a novel exploitation methodology to reliably leak plaintext enclave secrets from the CPU cache. We demonstrate our attacks by extracting full cryptographic keys from Intel's vetted architectural enclaves, and validate their correctness by launching rogue production enclaves and forging arbitrary local and remote attestation responses. The extracted remote attestation keys affect millions of devices.

Abstract: Physical systems with loss or gain have resonant modes that decay or grow exponentially with time. Whenever two such modes coalesce both in their resonant frequency and their rate of decay or growth, an 'exceptional point' occurs, giving rise to fascinating phenomena that defy our physical intuition. Particularly intriguing behaviour is predicted to appear when an exceptional point is encircled sufficiently slowly, such as a state-flip or the accumulation of a geometric phase. The topological structure of exceptional points has been experimentally explored, but a full dynamical encircling of such a point and the associated breakdown of adiabaticity have remained out of reach of measurement. Here we demonstrate that a dynamical encircling of an exceptional point is analogous to the scattering through a two-mode waveguide with suitably designed boundaries and losses. We present experimental results from a corresponding waveguide structure that steers incoming waves around an exceptional point during the transmission process. In this way, mode transitions are induced that transform this device into a robust and asymmetric switch between different waveguide modes. This work will enable the exploration of exceptional point physics in system control and state transfer schemes at the crossroads between fundamental research and practical applications.

Abstract: Brain metastases are associated with a dismal prognosis. Whether brain metastases harbor distinct genetic alterations beyond those observed in primary tumors is unknown. We performed whole-exome sequencing of 86 matched brain metastases, primary tumors, and normal tissue. In all clonally related cancer samples, we observed branched evolution, where all metastatic and primary sites shared a common ancestor yet continued to evolve independently. In 53% of cases, we found potentially clinically informative alterations in the brain metastases not detected in the matched primary-tumor sample. In contrast, spatially and temporally separated brain metastasis sites were genetically homogenous. Distal extracranial and regional lymph node metastases were highly divergent from brain metastases. We detected alterations associated with sensitivity to PI3K/AKT/mTOR, CDK, and HER2/EGFR inhibitors in the brain metastases. Genomic analysis of brain metastases provides an opportunity to identify potentially clinically informative alterations not detected in clinically sampled primary tumors, regional lymph nodes, or extracranial metastases. Significance: Decisions for individualized therapies in patients with brain metastasis are often made from primary-tumor biopsies. We demonstrate that clinically actionable alterations present in brain metastases are frequently not detected in primary biopsies, suggesting that sequencing of primary biopsies alone may miss a substantial number of opportunities for targeted therapy. Cancer Discov; 5(11); 1164–77. ©2015 AACR . See related commentary by Stricker and Arteaga, [p. 1124][1] . This article is highlighted in the In This Issue feature, [p. 1111][2] [1]: /lookup/volpage/5/1124?iss=11 [2]: /lookup/volpage/5/1111?iss=11

Abstract: It is always well believed that modeling relationships between objects would be helpful for representing and eventually describing an image Nevertheless, there has not been evidence in support of the idea on image description generation In this paper, we introduce a new design to explore the connections between objects for image captioning under the umbrella of attention-based encoder-decoder framework Specifically, we present Graph Convolutional Networks plus Long Short-Term Memory (dubbed as GCN-LSTM) architecture that novelly integrates both semantic and spatial object relationships into image encoder Technically, we build graphs over the detected objects in an image based on their spatial and semantic connections The representations of each region proposed on objects are then refined by leveraging graph structure through GCN With the learnt region-level features, our GCN-LSTM capitalizes on LSTM-based captioning framework with attention mechanism for sentence generation Extensive experiments are conducted on COCO image captioning dataset, and superior results are reported when comparing to state-of-the-art approaches More remarkably, GCN-LSTM increases CIDEr-D performance from 1201% to 1287% on COCO testing set

Abstract: During the last two decades, Internet-delivered cognitive behavior therapy (ICBT) has been tested in hundreds of randomized controlled trials, often with promising results. However, the control gro...

Abstract: BackgroundWhether patients with early-stage oral cancers should be treated with elective neck dissection at the time of the primary surgery or with therapeutic neck dissection after nodal relapse has been a matter of debate. MethodsIn this prospective, randomized, controlled trial, we evaluated the effect on survival of elective node dissection (ipsilateral neck dissection at the time of the primary surgery) versus therapeutic node dissection (watchful waiting followed by neck dissection for nodal relapse) in patients with lateralized stage T1 or T2 oral squamous-cell carcinomas. Primary and secondary end points were overall survival and disease-free survival, respectively. ResultsBetween 2004 and 2014, a total of 596 patients were enrolled. As prespecified by the data and safety monitoring committee, this report summarizes results for the first 500 patients (245 in the elective-surgery group and 255 in the therapeutic-surgery group), with a median follow-up of 39 months. There were 81 recurrences and 50 ...

Abstract: Objectives To describe a national cohort of pregnant women admitted to hospital with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection in the UK, identify factors associated with infection, and describe outcomes, including transmission of infection, for mothers and infants. Design Prospective national population based cohort study using the UK Obstetric Surveillance System (UKOSS). Setting All 194 obstetric units in the UK. Participants 427 pregnant women admitted to hospital with confirmed SARS-CoV-2 infection between 1 March 2020 and 14 April 2020. Main outcome measures Incidence of maternal hospital admission and infant infection. Rates of maternal death, level 3 critical care unit admission, fetal loss, caesarean birth, preterm birth, stillbirth, early neonatal death, and neonatal unit admission. Results The estimated incidence of admission to hospital with confirmed SARS-CoV-2 infection in pregnancy was 4.9 (95% confidence interval 4.5 to 5.4) per 1000 maternities. 233 (56%) pregnant women admitted to hospital with SARS-CoV-2 infection in pregnancy were from black or other ethnic minority groups, 281 (69%) were overweight or obese, 175 (41%) were aged 35 or over, and 145 (34%) had pre-existing comorbidities. 266 (62%) women gave birth or had a pregnancy loss; 196 (73%) gave birth at term. Forty one (10%) women admitted to hospital needed respiratory support, and five (1%) women died. Twelve (5%) of 265 infants tested positive for SARS-CoV-2 RNA, six of them within the first 12 hours after birth. Conclusions Most pregnant women admitted to hospital with SARS-CoV-2 infection were in the late second or third trimester, supporting guidance for continued social distancing measures in later pregnancy. Most had good outcomes, and transmission of SARS-CoV-2 to infants was uncommon. The high proportion of women from black or minority ethnic groups admitted with infection needs urgent investigation and explanation. Study registration ISRCTN 40092247.

Abstract: Estimates of biological age based on DNA methylation patterns, often referred to as "epigenetic age", "DNAm age", have been shown to be robust biomarkers of age in humans. We previously demonstrated that independent of chronological age, epigenetic age assessed in blood predicted all-cause mortality in four human cohorts. Here, we expanded our original observation to 13 different cohorts for a total sample size of 13,089 individuals, including three racial/ethnic groups. In addition, we examined whether incorporating information on blood cell composition into the epigenetic age metrics improves their predictive power for mortality. All considered measures of epigenetic age acceleration were predictive of mortality (p≤8.2x10-9), independent of chronological age, even after adjusting for additional risk factors (p<5.4x10-4), and within the racial/ethnic groups that we examined (non-Hispanic whites, Hispanics, African Americans). Epigenetic age estimates that incorporated information on blood cell composition led to the smallest p-values for time to death (p=7.5x10-43). Overall, this study a) strengthens the evidence that epigenetic age predicts all-cause mortality above and beyond chronological age and traditional risk factors, and b) demonstrates that epigenetic age estimates that incorporate information on blood cell counts lead to highly significant associations with all-cause mortality.

Abstract: Over the past few decades, metal nanoparticles less than 100 nm in diameter have made a substantial impact across diverse biomedical applications, such as diagnostic and medical devices, for personalized healthcare practice. In particular, silver nanoparticles (AgNPs) have great potential in a broad range of applications as antimicrobial agents, biomedical device coatings, drug-delivery carriers, imaging probes, and diagnostic and optoelectronic platforms, since they have discrete physical and optical properties and biochemical functionality tailored by diverse size- and shape-controlled AgNPs. In this review, we aimed to present major routes of synthesis of AgNPs, including physical, chemical, and biological synthesis processes, along with discrete physiochemical characteristics of AgNPs. We also discuss the underlying intricate molecular mechanisms behind their plasmonic properties on mono/bimetallic structures, potential cellular/microbial cytotoxicity, and optoelectronic property. Lastly, we conclude this review with a summary of current applications of AgNPs in nanoscience and nanomedicine and discuss their future perspectives in these areas.

Abstract: Spin–orbit coupling (SOC) describes the relativistic interaction between the spin and momentum degrees of freedom of electrons, and is central to the rich phenomena observed in condensed matter systems. In recent years, new phases of matter have emerged from the interplay between SOC and low dimensionality, such as chiral spin textures and spin-polarized surface and interface states. These low-dimensional SOC-based realizations are typically robust and can be exploited at room temperature. Here we discuss SOC as a means of producing such fundamentally new physical phenomena in thin films and heterostructures. We put into context the technological promise of these material classes for developing spin-based device applications at room temperature. The interplay between spin–orbit coupling and two-dimensionality has led to the emergence of new phases of matter, such as spin-polarized surface states in topological insulators, interfacial chiral spin interactions, and magnetic skyrmions in thin films, with great potential for spin-based devices. Substantial progress in the past decade in the fabrication and modelling of atomically precise interfaces and surfaces has led to the discovery of many electronic effects that are of interest for practical devices with novel functionalities. Christos Panagopoulos and colleagues review various such effects—and their technological potential—with a focus on the role of spin–orbit coupling, the fundamental interaction between the spin and charge degrees of freedom of an electron. Spin–orbit coupling can affect the electronic properties of materials at reduced dimensions, and the authors discuss the basic principles for understanding and engineering interfaces and surfaces in which spin–orbit coupling is harnessed. Examples are structures based on topological insulators where spin currents are generated or converted and magnetic layers where spin–orbit coupling leads to spin textures that can be controlled.

Abstract: In this paper, the efficient deployment and mobility of multiple unmanned aerial vehicles (UAVs), used as aerial base stations to collect data from ground Internet of Things (IoT) devices, are investigated. In particular, to enable reliable uplink communications for the IoT devices with a minimum total transmit power, a novel framework is proposed for jointly optimizing the 3D placement and the mobility of the UAVs, device-UAV association, and uplink power control. First, given the locations of active IoT devices at each time instant, the optimal UAVs’ locations and associations are determined. Next, to dynamically serve the IoT devices in a time-varying network, the optimal mobility patterns of the UAVs are analyzed. To this end, based on the activation process of the IoT devices, the time instances at which the UAVs must update their locations are derived. Moreover, the optimal 3D trajectory of each UAV is obtained in a way that the total energy used for the mobility of the UAVs is minimized while serving the IoT devices. Simulation results show that, using the proposed approach, the total-transmit power of the IoT devices is reduced by 45% compared with a case, in which stationary aerial base stations are deployed. In addition, the proposed approach can yield a maximum of 28% enhanced system reliability compared with the stationary case. The results also reveal an inherent tradeoff between the number of update times, the mobility of the UAVs, and the transmit power of the IoT devices. In essence, a higher number of updates can lead to lower transmit powers for the IoT devices at the cost of an increased mobility for the UAVs.

Abstract: Soil organisms are an integral component of ecosystems, but their activities receive little recognition in agricultural management strategies. Here we synthesize the potential of soil organisms to enhance ecosystem service delivery and demonstrate that soil biodiversity promotes multiple ecosystem functions simultaneously (i.e., ecosystem multifunctionality). We apply the concept of ecological intensification to soils and we develop strategies for targeted exploitation of soil biological traits. We compile promising approaches to enhance agricultural sustainability through the promotion of soil biodiversity and targeted management of soil community composition. We present soil ecological engineering as a concept to generate human land-use systems, which can serve immediate human needs while minimizing environmental impacts.

Abstract: Importance Sleeve gastrectomy is increasingly used in the treatment of morbid obesity, but its long-term outcome vs the standard Roux-en-Y gastric bypass procedure is unknown. Objective To determine whether there are differences between sleeve gastrectomy and Roux-en-Y gastric bypass in terms of weight loss, changes in comorbidities, increase in quality of life, and adverse events. Design, Setting, and Participants The Swiss Multicenter Bypass or Sleeve Study (SM-BOSS), a 2-group randomized trial, was conducted from January 2007 until November 2011 (last follow-up in March 2017). Of 3971 morbidly obese patients evaluated for bariatric surgery at 4 Swiss bariatric centers, 217 patients were enrolled and randomly assigned to sleeve gastrectomy or Roux-en-Y gastric bypass with a 5-year follow-up period. Interventions Patients were randomly assigned to undergo laparoscopic sleeve gastrectomy (n = 107) or laparoscopic Roux-en-Y gastric bypass (n = 110). Main Outcomes and Measures The primary end point was weight loss, expressed as percentage excess body mass index (BMI) loss. Exploratory end points were changes in comorbidities and adverse events. Results Among the 217 patients (mean age, 45.5 years; 72% women; mean BMI, 43.9) 205 (94.5%) completed the trial. Excess BMI loss was not significantly different at 5 years: for sleeve gastrectomy, 61.1%, vs Roux-en-Y gastric bypass, 68.3% (absolute difference, −7.18%; 95% CI, −14.30% to −0.06%;P = .22 after adjustment for multiple comparisons). Gastric reflux remission was observed more frequently after Roux-en-Y gastric bypass (60.4%) than after sleeve gastrectomy (25.0%). Gastric reflux worsened (more symptoms or increase in therapy) more often after sleeve gastrectomy (31.8%) than after Roux-en-Y gastric bypass (6.3%). The number of patients with reoperations or interventions was 16/101 (15.8%) after sleeve gastrectomy and 23/104 (22.1%) after Roux-en-Y gastric bypass. Conclusions and Relevance Among patients with morbid obesity, there was no significant difference in excess BMI loss between laparoscopic sleeve gastrectomy and laparoscopic Roux-en-Y gastric bypass at 5 years of follow-up after surgery. Trial Registration clinicaltrials.gov Identifier:NCT00356213

Abstract: State-of-the-art methods for zero-shot visual recognition formulate learning as a joint embedding problem of images and side information. In these formulations the current best complement to visual features are attributes: manuallyencoded vectors describing shared characteristics among categories. Despite good performance, attributes have limitations: (1) finer-grained recognition requires commensurately more attributes, and (2) attributes do not provide a natural language interface. We propose to overcome these limitations by training neural language models from scratch, i.e. without pre-training and only consuming words and characters. Our proposed models train end-to-end to align with the fine-grained and category-specific content of images. Natural language provides a flexible and compact way of encoding only the salient visual aspects for distinguishing categories. By training on raw text, our model can do inference on raw text as well, providing humans a familiar mode both for annotation and retrieval. Our model achieves strong performance on zero-shot text-based image retrieval and significantly outperforms the attribute-based state-of-the-art for zero-shot classification on the Caltech-UCSD Birds 200-2011 dataset.